Release 0.16.0

Adds table for storing camshaft analysis nodes

Makefile

@@ -1,7 +1,7 @@
 # cartodb/Makefile
 
 EXTENSION = cartodb
-EXTVERSION = 0.15.0
+EXTVERSION = 0.16.0
 
 SED = sed
 
@@ -65,6 +65,8 @@ UPGRADABLE = \
   0.14.3 \
   0.14.4 \
   0.15.0 \
+  0.15.1 \
+  0.16.0 \
   $(EXTVERSION)dev \
   $(EXTVERSION)next \
   $(END)

NEWS.md

@@ -1,5 +1,25 @@
-0.15.0 (2016-04-05)
+0.16.0 (2016-04-15)
+-------------------
+* Adds table for storing camshaft analysis nodes
+  [#237](https://github.com/CartoDB/cartodb-postgresql/pull/237)
 
+0.15.1 (2016-04-15)
+-------------------
+* Fix problems with org users in overviews functions
+  [#224](https://github.com/CartoDB/cartodb-postgresql/pull/224)
+* Add `_feature_count` to overviews
+  [#227](https://github.com/CartoDB/cartodb-postgresql/pull/227)
+* Change point clustering behaviour of overviews
+  [#228](https://github.com/CartoDB/cartodb-postgresql/pull/228)
+* Change default tolerance of overviews
+  [#230](https://github.com/CartoDB/cartodb-postgresql/pull/230)
+* Fix problem with aggregated numerical fields in overviews
+  [#233](https://github.com/CartoDB/cartodb-postgresql/pull/233)
+* Enhance aggregation of text fields in overviews
+  [#234]https://github.com/CartoDB/cartodb-postgresql/pull/234
+
+0.15.0 (2016-04-05)
+-------------------
 * New function CDB_CreateOverviewsWithToleranceInPixels that adds tolerance parameter for overview creation
   [#221](https://github.com/CartoDB/cartodb-postgresql/pull/221)
 * New default value for the overviews tolerance in pixels is 2 (used to be 7.5) (also in #221)
@@ -8,7 +28,7 @@
   [#220](https://github.com/CartoDB/cartodb-postgresql/pull/220)
 
 0.14.4 (2016-03-29)
-
+-------------------
 * Fix creating overviews for tables with boolean columns
   [#214](https://github.com/CartoDB/cartodb-postgresql/pull/214)
 * Fix tests for some systems [#215](https://github.com/CartoDB/cartodb-postgresql/pull/215)

scripts-available/CDB_AnalysisCatalog.sql

@@ -0,0 +1,24 @@
+-- Table to register analysis nodes from https://github.com/cartodb/camshaft
+CREATE TABLE IF NOT EXISTS
+cartodb.cdb_analysis_catalog (
+    -- md5 hex hash
+    node_id char(40) CONSTRAINT cdb_analysis_catalog_pkey PRIMARY KEY,
+    -- being json allows to do queries like analysis_def->>'type' = 'buffer'
+    analysis_def json NOT NULL,
+    -- can reference other nodes in this very same table, allowing recursive queries
+    input_nodes char(40) ARRAY NOT NULL DEFAULT '{}',
+    status TEXT NOT NULL DEFAULT 'pending',
+    CONSTRAINT valid_status CHECK (
+        status IN ( 'pending', 'waiting', 'running', 'canceled', 'failed', 'ready' )
+    ),
+    created_at timestamp with time zone NOT NULL DEFAULT now(),
+    -- should be updated when some operation was performed in the node
+    -- and anything associated to it might have changed
+    updated_at timestamp with time zone DEFAULT NULL,
+    -- should register last time the node was used
+    used_at timestamp with time zone NOT NULL DEFAULT now(),
+    -- should register the number of times the node was used
+    hits NUMERIC DEFAULT 0,
+    -- should register what was the last node using current node
+    last_used_from char(40)
+);

scripts-available/CDB_Overviews.sql

@@ -1,4 +1,24 @@
--- security definer
+-- Information about tables in a schema.
+-- If the schema name parameter is NULL, then tables from all schemas
+-- that may contain user tables are returned.
+-- For each table, the regclass, schema name and table name are returned.
+-- Scope: private.
+CREATE OR REPLACE FUNCTION _CDB_UserTablesInSchema(schema_name text DEFAULT NULL)
+RETURNS TABLE(table_regclass REGCLASS, schema_name TEXT, table_name TEXT)
+AS $$
+  SELECT
+    c.oid::regclass AS table_regclass,
+    n.nspname::text AS schema_name,
+    c.relname::text AS table_relname
+  FROM pg_class c
+  JOIN pg_namespace n ON n.oid = c.relnamespace
+  WHERE c.relkind = 'r'
+  AND c.relname NOT IN ('cdb_tablemetadata', 'spatial_ref_sys')
+  AND CASE WHEN schema_name IS NULL
+             THEN n.nspname NOT IN ('pg_catalog', 'information_schema', 'topology', 'cartodb')
+           ELSE n.nspname = schema_name
+           END;
+$$ LANGUAGE 'sql';
 
 -- Pattern that can be used to detect overview tables and Extract
 -- the intended zoom level from the table name.
@@ -120,7 +140,7 @@ BEGIN
     FOR row IN
         SELECT * FROM CDB_Overviews(reloid)
     LOOP
-        EXECUTE Format('DROP TABLE %I.%I;', schema_name, row.overview_table);
+        EXECUTE Format('DROP TABLE %s;', row.overview_table);
         RAISE NOTICE 'Dropped overview for level %: %', row.z, row.overview_table;
     END LOOP;
 END;
@@ -140,16 +160,15 @@ RETURNS TABLE(base_table REGCLASS, z integer, overview_table REGCLASS)
 AS $$
   DECLARE
     schema_name TEXT;
-    table_name TEXT;
+    base_table_name TEXT;
   BEGIN
-    -- TODO: review implementation of CDB_UserTables an suitability for this
-    SELECT * FROM _cdb_split_table_name(reloid) INTO schema_name, table_name;
+    SELECT * FROM _cdb_split_table_name(reloid) INTO schema_name, base_table_name;
     RETURN QUERY SELECT
       reloid AS base_table,
-      _CDB_OverviewTableZ(cdb_usertables) AS z,
-      ('"' || schema_name|| '"."' ||cdb_usertables || '"')::regclass AS overview_table
-      FROM CDB_UserTables()
-      WHERE _CDB_IsOverviewTableOf((SELECT relname FROM pg_class WHERE oid=reloid), cdb_usertables)
+      _CDB_OverviewTableZ(table_name) AS z,
+      table_regclass AS overview_table
+      FROM _CDB_UserTablesInSchema(schema_name)
+      WHERE _CDB_IsOverviewTableOf((SELECT relname FROM pg_class WHERE oid=reloid), table_name)
       ORDER BY z;
   END
 $$ LANGUAGE PLPGSQL;
@@ -168,11 +187,13 @@ RETURNS TABLE(base_table REGCLASS, z integer, overview_table REGCLASS)
 AS $$
   SELECT
     base_table::regclass AS base_table,
-    _CDB_OverviewTableZ(cdb_usertables) AS z,
-    ('"' || _cdb_schema_name(base_table::regclass) || '"."' || cdb_usertables || '"')::regclass AS overview_table
+    _CDB_OverviewTableZ(table_name) AS z,
+    table_regclass AS overview_table
     FROM
-      CDB_UserTables(), unnest(tables) base_table
-    WHERE _CDB_IsOverviewTableOf((SELECT relname FROM pg_class WHERE oid=base_table), cdb_usertables)
+      _CDB_UserTablesInSchema(), unnest(tables) base_table
+    WHERE
+      schema_name = _cdb_schema_name(base_table)
+      AND _CDB_IsOverviewTableOf((SELECT relname FROM pg_class WHERE oid=base_table), table_name)
     ORDER BY base_table, z;
 $$ LANGUAGE SQL;
 
@@ -200,11 +221,17 @@ AS $$
 
     BEGIN
       EXECUTE ext_query INTO ext;
-      EXCEPTION
+    EXCEPTION
         -- This is the typical ERROR: stats for "mytable" do not exist
         WHEN internal_error THEN
           -- Get stats and execute again
-          EXECUTE format('ANALYZE %1$I', reloid);
+          EXECUTE format('ANALYZE %1$s', reloid);
+
+          -- We check the geometry type in case the error is due to empty geometries
+          IF _CDB_GeometryTypes(reloid) IS NULL THEN
+            RETURN NULL;
+          END IF;
+
           EXECUTE ext_query INTO ext;
     END;
 
@@ -370,7 +397,7 @@ AS $$
 
     SELECT * FROM _cdb_split_table_name(reloid) INTO schema_name, table_name;
 
-    EXECUTE Format('DROP TABLE IF EXISTS %I.%I CASCADE;', schema_name.overview_rel);
+    EXECUTE Format('DROP TABLE IF EXISTS %I.%I CASCADE;', schema_name, overview_rel);
 
     -- Estimate number of rows
     SELECT reltuples, relpages FROM pg_class INTO STRICT class_info
@@ -384,16 +411,16 @@ AS $$
     ELSE
       num_samples := ceil(class_info.reltuples*fraction);
       EXECUTE Format('
-        CREATE TABLE %1$I AS SELECT * FROM %2$s
+        CREATE TABLE %4$I.%1$I AS SELECT * FROM %2$s
           WHERE ctid = ANY (
             ARRAY[
               (SELECT CDB_RandomTids(''%2$s'', %3$s))
             ]
           );
-      ', overview_rel, reloid, num_samples);
+      ', overview_rel, reloid, num_samples, schema_name);
     END IF;
 
-    RETURN overview_rel;
+    RETURN Format('%I.%I', schema_name, overview_rel)::regclass;
   END;
 $$ LANGUAGE PLPGSQL;
 
@@ -429,9 +456,12 @@ AS $$
 
       -- preserve the owner of the base table
       SELECT u.usename
-        FROM pg_catalog.pg_class c JOIN pg_catalog.pg_user u ON (c.relowner=u.usesysid)
-        WHERE c.relname = dataset::text
+        FROM pg_catalog.pg_class c
+          JOIN pg_catalog.pg_user u ON (c.relowner=u.usesysid)
+          JOIN pg_namespace n ON n.oid = c.relnamespace
+        WHERE c.relname = dataset_name::text AND n.nspname = dataset_scheme
         INTO table_owner;
+
       EXECUTE Format('ALTER TABLE IF EXISTS %s OWNER TO %I;', overview_table::text, table_owner);
 
       -- preserve the table privileges
@@ -485,6 +515,23 @@ BEGIN
 END
 $$ LANGUAGE PLPGSQL STABLE;
 
+-- Check if a column of a table is of an unlimited-length text type
+CREATE OR REPLACE FUNCTION _cdb_unlimited_text_column(reloid REGCLASS, col_name TEXT)
+RETURNS BOOLEAN
+AS $$
+  SELECT EXISTS (
+    SELECT  *
+      FROM information_schema.columns c, pg_class _tn, pg_namespace _sn
+      WHERE table_name = _tn.relname
+        AND table_schema = _sn.nspname
+        AND c.column_name = col_name
+        AND _tn.oid = reloid
+        AND _sn.oid = _tn.relnamespace
+        AND character_maximum_length IS NULL
+        AND c.data_type IN ('text', 'character varying', 'character')
+  );
+$$ LANGUAGE SQL STABLE;
+
 -- SQL Aggregation expression for a datase attribute
 -- Scope: private.
 -- Parameters
@@ -499,6 +546,9 @@ AS $$
 DECLARE
   column_type TEXT;
   qualified_column TEXT;
+  has_counter_column BOOLEAN;
+  feature_count TEXT;
+  total_feature_count TEXT;
 BEGIN
   IF table_alias <> '' THEN
     qualified_column := Format('%I.%I', table_alias, column_name);
@@ -508,19 +558,32 @@ BEGIN
 
   column_type := CDB_ColumnType(reloid, column_name);
 
+  SELECT EXISTS (
+    SELECT * FROM CDB_ColumnNames(reloid)  as colname WHERE colname = '_feature_count'
+  ) INTO has_counter_column;
+  IF has_counter_column THEN
+    feature_count := '_feature_count';
+    total_feature_count := 'SUM(_feature_count)';
+  ELSE
+    feature_count := '1';
+    total_feature_count := 'count(*)';
+  END IF;
+
   CASE column_type
-  WHEN 'double precision', 'real', 'integer', 'bigint' THEN
-    RETURN Format('AVG(%s)::' || column_type, qualified_column);
-  WHEN 'text' THEN
-    -- TODO: we could define a new aggregate function that returns distinct
-    -- separated values with a limit, adding ellipsis if more values existed
-    -- e.g. with '/' as separator and a limit of three:
-    --     'A', 'B', 'A', 'C', 'D' => 'A/B/C/...'
-    -- Other ideas: if value is unique then use it, otherwise use something
-    -- like '*' or '(varies)' or '(multiple values)', or NULL
-    -- Using 'string_agg(' || qualified_column || ',''/'')'
-    -- here causes
-    RETURN 'CASE count(*) WHEN 1 THEN MIN(' || qualified_column || ') ELSE NULL END::' || column_type;
+  WHEN 'double precision', 'real', 'integer', 'bigint', 'numeric' THEN
+    IF column_name = '_feature_count' THEN
+      RETURN 'SUM(_feature_count)';
+    ELSE
+      RETURN Format('SUM(%s*%s)/%s::' || column_type, qualified_column, feature_count, total_feature_count);
+    END IF;
+  WHEN 'text', 'character varying', 'character' THEN
+    IF _cdb_unlimited_text_column(reloid, column_name) THEN
+      -- TODO: this should not be applied to columns containing largish text;
+      -- it is intended only to short names/identifiers
+      RETURN  'CASE WHEN count(distinct ' || qualified_column || ') = 1 THEN MIN(' || qualified_column || ') WHEN ' || total_feature_count || ' < 5 THEN string_agg(distinct ' || qualified_column || ','' / '') ELSE ''*'' END::' || column_type;
+    ELSE
+      RETURN 'CASE count(*) WHEN 1 THEN MIN(' || qualified_column || ') ELSE NULL END::' || column_type;
+    END IF;
   WHEN 'boolean' THEN
     RETURN 'CASE count(*) WHEN 1 THEN BOOL_AND(' || qualified_column || ') ELSE NULL END::' || column_type;
   ELSE
@@ -596,12 +659,12 @@ AS $$
     gtypes TEXT[];
     schema_name TEXT;
     table_name TEXT;
+    point_geom TEXT;
   BEGIN
     SELECT _CDB_GeometryTypes(reloid) INTO gtypes;
-    IF array_upper(gtypes, 1) <> 1 OR gtypes[1] <> 'ST_Point' THEN
+    IF gtypes IS NULL OR array_upper(gtypes, 1) <> 1 OR gtypes[1] <> 'ST_Point' THEN
       -- This strategy only supports datasets with point geomety
       RETURN NULL;
-      RETURN 'x';
     END IF;
 
     --TODO: check applicability: geometry type, minimum number of points...
@@ -610,7 +673,7 @@ AS $$
 
     -- Grid size in pixels at Z level overview_z
     IF grid_px IS NULL THEN
-      grid_px := 7.5;
+      grid_px := 1.0;
     END IF;
 
     SELECT * FROM _cdb_split_table_name(reloid) INTO schema_name, table_name;
@@ -627,19 +690,17 @@ AS $$
       aggr_attributes := aggr_attributes || ', ';
     END IF;
 
+    point_geom = Format('ST_SetSRID(ST_MakePoint(gx*%1$s + %2$s, gy*%1$s + %2$s), 3857)', grid_m, grid_m/2);
+
     -- compute the resulting columns in the same order as in the base table
-    -- cartodb_id,
-    -- ST_Transform(ST_SetSRID(ST_MakePoint(sx/n, sy/n), 3857), 4326) AS the_geom,
-    -- ST_SetSRID(ST_MakePoint(sx/n, sy/n), 3857) AS the_geom_webmercator
-    -- %4$s
     WITH cols AS (
       SELECT
         CASE c
         WHEN 'cartodb_id' THEN 'cartodb_id'
         WHEN 'the_geom' THEN
-          'ST_Transform(ST_SetSRID(ST_MakePoint(sx/n, sy/n), 3857), 4326) AS the_geom'
+          Format('ST_Transform(%s, 4326) AS the_geom', point_geom)
         WHEN 'the_geom_webmercator' THEN
-           'ST_SetSRID(ST_MakePoint(sx/n, sy/n), 3857) AS the_geom_webmercator'
+           Format('%s AS the_geom_webmercator', point_geom)
         ELSE c
         END AS column
         FROM CDB_ColumnNames(reloid) c
@@ -648,6 +709,10 @@ AS $$
       SELECT * FROM cols
     ) AS s INTO columns;
 
+    IF NOT columns LIKE '%_feature_count%' THEN
+      columns := columns || ', n AS _feature_count';
+    END IF;
+
     EXECUTE Format('DROP TABLE IF EXISTS %I.%I CASCADE;', schema_name, overview_rel);
 
     -- Now we cluster the data using a grid of size grid_m
@@ -655,13 +720,11 @@ AS $$
     -- If we had a selected numeric attribute of interest we could use it
     -- as a weight for the average coordinates.
     EXECUTE Format('
-      CREATE TABLE %3$I AS
+      CREATE TABLE %7$I.%3$I AS
          WITH clusters AS (
            SELECT
              %5$s
              count(*) AS n,
-             SUM(ST_X(f.the_geom_webmercator)) AS sx,
-             SUM(ST_Y(f.the_geom_webmercator)) AS sy,
              Floor(ST_X(f.the_geom_webmercator)/%2$s)::int AS gx,
              Floor(ST_Y(f.the_geom_webmercator)/%2$s)::int AS gy,
              MIN(cartodb_id) AS cartodb_id
@@ -669,9 +732,9 @@ AS $$
           GROUP BY gx, gy
          )
          SELECT %6$s FROM clusters
-    ', reloid::text, grid_m, overview_rel, attributes, aggr_attributes, columns);
+    ', reloid::text, grid_m, overview_rel, attributes, aggr_attributes, columns, schema_name);
 
-    RETURN overview_rel;
+    RETURN Format('%I.%I', schema_name, overview_rel)::regclass;
   END;
 $$ LANGUAGE PLPGSQL;
 
@@ -693,7 +756,7 @@ DECLARE
   tolerance_px FLOAT8;
 BEGIN
   -- Use the default tolerance
-  tolerance_px := 2.0;
+  tolerance_px := 1.0;
   RETURN CDB_CreateOverviewsWithToleranceInPixels(reloid, tolerance_px, refscale_strategy, reduce_strategy);
 END;
 $$ LANGUAGE PLPGSQL;
@@ -710,10 +773,15 @@ DECLARE
   overview_z integer;
   overview_tables REGCLASS[];
   overviews_step integer := 1;
+  has_counter_column boolean;
 BEGIN
   -- Determine the referece zoom level
   EXECUTE 'SELECT ' || quote_ident(refscale_strategy::text) || Format('(''%s'', %s);', reloid, tolerance_px) INTO ref_z;
 
+  IF ref_z < 0 OR ref_z IS NULL THEN
+    RETURN NULL;
+  END IF;
+
   -- Determine overlay zoom levels
   -- TODO: should be handled by the refscale_strategy?
   overview_z := ref_z - 1;
@@ -735,6 +803,17 @@ BEGIN
     SELECT array_append(overview_tables, base_rel) INTO overview_tables;
   END LOOP;
 
+  IF overview_tables IS NOT NULL AND array_length(overview_tables, 1) > 0 THEN
+    SELECT EXISTS (
+      SELECT * FROM CDB_ColumnNames(reloid)  as colname WHERE colname = '_feature_count'
+    ) INTO has_counter_column;
+    IF NOT has_counter_column THEN
+      EXECUTE Format('
+        ALTER TABLE %s ADD COLUMN _feature_count integer DEFAULT 1;
+      ', reloid);
+    END IF;
+  END IF;
+
   RETURN overview_tables;
 END;
 $$ LANGUAGE PLPGSQL;

scripts-enabled/260-CDB_AnalysisCatalog.sql

@@ -0,0 +1 @@
+../scripts-available/CDB_AnalysisCatalog.sql

test/CDB_OverviewsTest_expect

@@ -9,35 +9,30 @@ SELECT 1114
 
 
 
-{_vovw_3_base_bare_t,_vovw_2_base_bare_t,_vovw_1_base_bare_t,_vovw_0_base_bare_t}
-113
+{_vovw_2_base_bare_t,_vovw_1_base_bare_t,_vovw_0_base_bare_t}
+126
 number,int_number,name,start
-AVG(number)::double precision AS number,AVG(int_number)::integer AS int_number,CASE count(*) WHEN 1 THEN MIN(name) ELSE NULL END::text AS name,CASE count(*) WHEN 1 THEN MIN(start) ELSE NULL END::date AS start
-AVG(tab.number)::double precision AS number,AVG(tab.int_number)::integer AS int_number,CASE count(*) WHEN 1 THEN MIN(tab.name) ELSE NULL END::text AS name,CASE count(*) WHEN 1 THEN MIN(tab.start) ELSE NULL END::date AS start
-{_vovw_3_base_t,_vovw_2_base_t,_vovw_1_base_t,_vovw_0_base_t}
-113
+SUM(number*1)/count(*)::double precision AS number,SUM(int_number*1)/count(*)::integer AS int_number,CASE WHEN count(distinct name) = 1 THEN MIN(name) WHEN count(*) < 5 THEN string_agg(distinct name,' / ') ELSE '*' END::text AS name,CASE count(*) WHEN 1 THEN MIN(start) ELSE NULL END::date AS start
+SUM(tab.number*1)/count(*)::double precision AS number,SUM(tab.int_number*1)/count(*)::integer AS int_number,CASE WHEN count(distinct tab.name) = 1 THEN MIN(tab.name) WHEN count(*) < 5 THEN string_agg(distinct tab.name,' / ') ELSE '*' END::text AS name,CASE count(*) WHEN 1 THEN MIN(tab.start) ELSE NULL END::date AS start
+{_vovw_2_base_t,_vovw_1_base_t,_vovw_0_base_t}
+126
 
-{_vovw_3_column_types_t,_vovw_2_column_types_t,_vovw_1_column_types_t,_vovw_0_column_types_t}
+{_vovw_2_column_types_t,_vovw_1_column_types_t,_vovw_0_column_types_t}
 (base_t,0,_vovw_0_base_t)
 (base_t,1,_vovw_1_base_t)
 (base_t,2,_vovw_2_base_t)
-(base_t,3,_vovw_3_base_t)
 (base_t,0,_vovw_0_base_t)
 (base_t,1,_vovw_1_base_t)
 (base_t,2,_vovw_2_base_t)
-(base_t,3,_vovw_3_base_t)
 (base_bare_t,0,_vovw_0_base_bare_t)
 (base_bare_t,1,_vovw_1_base_bare_t)
 (base_bare_t,2,_vovw_2_base_bare_t)
-(base_bare_t,3,_vovw_3_base_bare_t)
 (base_t,0,_vovw_0_base_t)
 (base_t,1,_vovw_1_base_t)
 (base_t,2,_vovw_2_base_t)
-(base_t,3,_vovw_3_base_t)
 (column_types_t,0,_vovw_0_column_types_t)
 (column_types_t,1,_vovw_1_column_types_t)
 (column_types_t,2,_vovw_2_column_types_t)
-(column_types_t,3,_vovw_3_column_types_t)
 
 
 

test/extension/run_at_cartodb_schema.sql

@@ -3,4 +3,5 @@ SET SCHEMA 'cartodb';
 \i scripts-available/CDB_TableMetadata.sql
 \i scripts-available/CDB_ColumnNames.sql
 \i scripts-available/CDB_ColumnType.sql
+\i scripts-available/CDB_AnalysisCatalog.sql
 SET SCHEMA 'public';

test/extension/test.sh

@@ -563,6 +563,13 @@ test_extension|public|"local-table-with-dashes"'
     DATABASE=fdw_target tear_down_database
 }
 
+function test_cdb_catalog_basic_node() {
+    DEF="'{\"type\":\"buffer\",\"source\":\"b2db66bc7ac02e135fd20bbfef0fdd81b2d15fad\",\"radio\":10000}'"
+    sql postgres "INSERT INTO cartodb.cdb_analysis_catalog (node_id, analysis_def) VALUES ('1bbc4c41ea7c9d3a7dc1509727f698b7', ${DEF}::json)"
+    sql postgres "SELECT status from cartodb.cdb_analysis_catalog where node_id = '1bbc4c41ea7c9d3a7dc1509727f698b7'" should 'pending'
+    sql postgres "DELETE FROM cartodb.cdb_analysis_catalog"
+}
+
 #################################################### TESTS END HERE ####################################################
 
 run_tests $@